Absorber Design in Sour Natural Gas Treatment Plants: Impact of Process Variables on Operation and Economics

Absorberentwurf ftir Anlagen zur Behandlung von sauren Erdgasen Prozessparameter auf Betriebsfihrung und Prozesswirtschaftlichkeit
P. M. M. BLAUWHOFF*, B. KAMPHUIS, W. P. M. VAN SWAAIJ and K. R. WESTERTERP

EinUuss

Chemical Reaction Engineering Laboratories, Department of Chemical Engineering, Twente University of Technology, P.O. Box 217, 7500 AE Enschede (The Netherlands)
(Received November 1, 1982)

Abstract
Two models of absorber have been developed which describe the absorption of H2S and CO2 from natural gases by aqueous di-isopropanolamine (DIPA) or methyl-di-ethanolamine (MDEA) solutions. In these models mass transfer, reaction and equilibrium processes as they prevail in conventional tray absorbers and in cascades of trickle bed reactors are incorporated. Owing to the better mass transfer characteristics of the latter type of absorber, i.e. the larger ratio between the gas phase and liquid phase mass transfer coefficients, kp/kp, higher selectivities for the absorption of HzS from sour natural gases are realized. The influence of variation of a number of operation and design parameters on tray absorber performance, HaS selectivity and solvent flowrate is demonstrated. The economics of the above type of absorbers together with a solvent regenerator, sulfur recovery unit and tail-gas unit are explained in detail. From the point of view of the economics trickle bed absorbers are very attractive owing to lower investment costs and higher selectivities, which result in lower operating costs than for tray absorbers under identical conditions.

Kurzfassung
Zwei Absorbermodelle wurden entwickelt zur Beschreibung der Absorption von Schwefelwasserstoff und Kohlendioxid aus Erdgasen mit Hilfe wisseriger Di-isopropanolamin (DIPA) oder Methyl-di-iithanolamin (MDEA) Liisungen. In diesen Modellen slnd Stoffaustauschvorgange, Reaktionen und Gleichgewichte verarbeitet wie sie in konventionellen Bodenkolonnen und in Kaskaden von Trickle-bedabsorbern vorliegen. Durch bessere Stoffaustauschcharakteristiken, d.h. einem grosserem Verhsltnis zwischen Gasphase- und Fltissigkeitsstoffaustauschkoeffizienten (k&a), kann im letzteren Absorbertyp eine hohere Selektivitat fiir die Absorption des Schwefelwasserstoffs aus sauren Erdgasen erzielt werden. Der Einfluss von Varlationen mehrerer Betriebs- und Entwurfparameter auf die Bodenkolonnenleistung, HZSSelektivitat und den Losungsmitteldurchsatz werden demonstriert. Die Wirtschaftlichkeit der genannten Absorbertypen in Verbindung mit einem Lijsungsmittelregenerator, einer Schwefelrtickgewinnungsanlage und einer Tail-gas Anlage werden im Detail erlautert. Aus dem Gesichtspunkt der Wirtschaftlichkeit sind Trickle-bedabsorber sehr attraktiv durch geringere Investitionen und hohere Selektivitaten, die in niedrigeren Betriebskosten resultieren als bei Bodenkolonnen identischer Leistung.

Synopse Zur Beschreibung der Hochdruckabswlpttin (70 bar) van Schwefehvasserstoff und Kohlendioxid aus Erdgasen mit Hilfe wtissetigen Di-isopropanolamin (DIPA) und Methyl-di4thanolamin (MDEA) Ldsungen wurden zwei Absorbermodelle entwickelt. Das erste Model1 beschreibt relevante physikalische und chemische Prozesse, die in
*Present address: Koninklijke/Shell-Laboratorium Amsterdam, P.O. BOX 3003, 1003 AA Amsterdam, The Netherlands.

konventionellen Bodenkolonnen stattfinden (F&. 2). Das zweite Modell beschreibt diese Prozesse in einer Trickle-bedabsorberkaskade (Fig. 3). Die Kolonne wind Boden fir Boden durchgerechnet. Fur j&en ideal vermischt angenommenen Boden werden die physikalischen und chemischen Gleichgewichte, die Reaktionsund Stoffiibertragungsgeschwind&keiten. die Nicht-idea&tit der Gasphase und die Warmeeffekte bei den herschenden Bedingungen berechnet. Die Gleichgewichte werden mit einen relativ einfachen, aber effektiven Model1 ermittelt (II]. Die Stoffbertragungsl-25 0 Elsevier Sequoia/Printed in The Netherlands

0255-2701/85/$3.30

Chem, Eng. Process., 19 (1985)

geschwindigkeiten werden sowohl mit einer analytischen Losung, die Wechselwirkung zwischen und Reversibilitat der Flussigkeitsreaktionen des Schwefelwasserstoffs und Kohlendioxids mit Alkanokamin vernachl&igt. als such numerisch unter Beriicksichtigung dieser Effekte, berechnet f 131. Eine Modifikation der Redlich-Kwong Zustandsgleichung nach Soave beschreibt die Nichtkiealitat der Gasphase. Bei der Berechnung der Tricklebedabsorberkaskade wird jeder Gleichstromabsorber als Reihe ideal vermischter Sektionen angenommen. Jede Sektion wird analog zum Boden berechnet. Die beiden Methoden zur Beschreibung der Stoffubertragung (analytisch und numerisch) werden an Hand des Modells der Bodenkolonne verglichen. Unter normalen Bedingungen sind die Unterschiede zwischen beiden Modellen vernachkissigbar (Fig. 5-8). Bei extremen Bedingungen ist jedoch das numertsche Model1 notwendig. Im grossten Teil derKolonneistfiir Schwefelwasserstoff die Stoffibertragungsgeschwindigkeit in der die Gasphase limitiert, wahrend fur Kohlendioxid Limitierung in der Flussigkeit liegt. Ein Vergleich der Alkanoknninl&ungen zeigt, dass das tertiare Amin, MDEA, nur unwesentlich selektiver ist ah das sekundare DIPA. Fur betie Amine liegt die Selektivitat, q (definiert durch Gl. (33)), nahe am theoretischen Minimum. Aus den Berechnungen zetgt sich weiter, dass der Betnebsdruck der Kolonne die Selektivitat q und den Ldsungsmitteldurchsatz Ga nicht beeinflusst (Fig. 9). Die anftingliche Beladung des regenerierten Ldsungsmittels mit HzS und COa hat einen ziemlich starken Einjluss auf 77und insbesondere auf $Q (Fig. IO). Die Endbeladung des Losungsmittels steuert die Selektivitat kaum, hat aber einen grossen Einfluss auf den Losungsmitteldurchsatz (Fip. II). Bei hohen Aminkonzentrationen und deshalb niedrigen tisungsmitteldurchsatzen, f%hrt die Warmeentwicklung zu hohen Temperaturen unten in der Kolonne und kann Kohlendioxid desorbieren (Ftg 13-15). Nach den Modellberechnungen zeigt die Tricklebedabsorberkaskade eine deutlich bessere Selektivitat und eirten giinstigeren Losungsmitteldurchsatz als die Bodenkolonne. Der Einjluss der Selektivitat der Hochdruckabsorber auf die Betnebskosten einer vollst&digen Gasbehandlungsanlage, d.h. ein Hochdnrckabsorber mit Ldsungsmittelregenerator, eine Schwefeltickgewinnungsanlage und eine Tail-gas Anlage (Fig. 1) wird mit Hilfe mathematischer Modelle der Anlagen berechnet. Mit steigender Selektivitat der Hochdruckabsolption sinken die Gasbehandhngskosten betrtihtlich (Fig. 20). Die Abhcingigkeit der Betrtebskosten vom Schwefebreis und von rekativen Energiekosten wird aufgezeigt (Fig. 22 und 23). Trickle-bedabsorberkaskaden sind sehr attraktiv durch niedrigere Investittons und Betriebskosten ah Bodenkolonnen tiientischer Leistung.

introduction
Removal of the acid components H2S and CO* from sour natural gases by means of alkanolamine solutions

is studied in this paper. The industrially most important alkanolamines for gas treatment operations are monoethanolamine (MEA), di-ethanolamine (DEA), diisopropanolamine (DIPA) and methyl-di-ethanolamine (MDEA) [l, 21. In general these amines are used in aqueous solutions but for specific applications combined solvents can be more suitable, e.g. water and sulfolane in the Shell Sulfinol process [ 1,2]. In view of the high energy consumption in sour natural gas treatment plants, there is considerable incentive for the development of-even slightly-more efficient processes [3,4]. Large savings in operating and investment costs can be obtained by the selective absorption of HsS alone from HsS and CO2 containing gases. Even in liquid natural gas production, where the CO* ultimately also has to be removed in order to avoid plugging of cryogenic equipment, selective absorption of H2S is economically very attractive using sophisticated treatment schemes as described, for example, by McEwan and Marmin [4]. The selective absorption of HsS from a sour gas containing CO2 also offers a number of economical advantages: - it reduces the solvent circulation rate and, as a consequence, the regeneration costs and investment in equipment are reduced; - it increases the H2S/C02 ratio in the off-gas to the sulfur recovery unit, which reduces the dimensions of and investment in the sulfur recovery and tailgas units. The H2S selectivity in gas treatment plants using alkanolamine solvents depends largely on three factors: _ the mass transfer properties of the high pressure absorber; - the (chemical) equilibria in the HzS-COz-amine system; _ the kinetics of the reactions between H,S/COs and the amines. In order to study the effect of these factors on the economics of the treatment operation, we developed a set of mathematical models describing a complete sour natural gas treatment plant. The flow scheme of the plant considered is shown in Fig. 1 and consists of three basic units: - a high pressure absorber-regenerator unit (ARU); - a Claus sulfur recovery unit (SRU); - a Shell Claus off-gas treater (SCOT) tail-gas unit (TGU). In order to limit the complexity of the calculations and consequently the computer costs involved, we used this relatively simple flowsheet as the basis for this study. We realize, of course, that more sophisticated treatment schemes, involving, for example, cascaded and splitstream solvent flows, are described in the literature and used in commercial operations [ 1,4,5]. As will be shown, the operation of the high pressure absorber is a key parameter in the overall process economics. For this reason the emphasis in this study is put on the relation between the absorber design and the plant performance. Tray column absorbers are used nowadays in large-scale treatment plants. Therefore, the performance of this type of absorber is analysed

ABSORBER

-REGENERATOR

UNIT

lARV) ______-_-7-I ------

SULPHUR

RECOVERY

UNIT

ISRU)

---------__-_____.________,

absorber TAIL-GAS UNIT ITGUI

Fig. 1. Flow scheme of the treatment plant.

for different operating conditions. Moreover, the properties of a potentially new absorber type for these operations-a cascade of trickle beds-will also be evaluated. Modifications in process set-ups like those mentioned above can be judged and evaluated only by their impact on the overall process economics. Therefore we incorporated in this study a comparisonoftheprocess economics for a number of absorber options. This is carried out on the basis of investment and cost calculations for each of these options, resulting in a cost for the treatment operation per Nm3 of natural gas. For a study like this, a vast amount of data is needed. Wherever possible, these data have been extracted from the open literature. Since this study covers a subject of commercial importance, information on the process economics [6, 71, absorber [6,8] and regenerator design [7] is very scarce. Nevertheless, we are convinced that the results shown in this paper give a good understanding of the impact of absorber design and operation on selectivity and the overall process economics of gas treatment. For several years now, the fundamentals of (alkanolamine) gas treatment have been one of the subjects of the research programme in our laboratories of Chemical Reaction Engineering at the Twente University of Technology. The emphasis in our research is on improving the selectivity of the absorption of H2S from H2S and CO2 containing gases like natural gas by means of

the mass transfer properties in absorbers. To this end fundamental studies were executed on the following items: a general description of the rate of the reaction between CO2 and aqueous alkanolamines [9, lo] ; the development of a mathematical model for the equilibria in the HsS-COs-amine-water system [9,11,121; the mathematical description of the simultaneous mass transfer of H2S and CO2 from the gas phase and the complex reversible liquid phase reactions [9, 131. In the work presented here we apply the knowledge acquired in previous studies [9-131 to the design of a complete treatment plant and also incorporate an evaluation of the economics of the amine treatment process. A summary of the reaction scheme and the pertinent kinetic data is given in Table A.

Design Considerations
General The flowsheet of the treatment plant studied is shown in Fig. 1. Using aqueous alkanolamine solutions the temperatures in the regenerator are relatively low. Consequently the amine losses due to (high temperature) degradation reactions are low and are assumed to be balanced by the amine make-up for compensation of pump losses and leakages. For this reason an amine

4 TABLE A. Reaction 1. Equilibrium RiRsNCOOfor DIPA: 2. Reaction DIPA: data: + Hz0 5 RiRsNH + HCOs191 scheme and data for HsS + CO2 absorption in DIPA

K = 1.01 X104 exp(-3173.9/T) rate expressions: r= ki,r[COs](DIPA]

[9,141

where /cl,1 = 4.09 X log exp(-4808/T)

MDEA: where kr,, r=klJC02][MDEA] = 1.02 X 1Oa exp(-4808/T) 19,101

3. Liquid phase viscosities: &DIPA) = Il.244 - 5.47 X 10U4[DIPA] + 3.668 X 10-7[DIPA]Z} exp(4549/T ~ 14.31) CP CP [91 I91

/.i(MDEA) = (0.7747 Gas Interface

- 6.941 X 10-5[MDEA] Liquid

+ 1.4934 X 10-7[MDEA]Z}

exp(2977/T

- 9.36)

H+ + COs2-

Irreversible degradation (oxazolidon, urea)

HC03-

JI
+

+ RsNH =

Hz0

+ RsNCOO-

T
I

co2g

C&Q

2RsNH

RsNHs+ + RsNCOO-

/Hz0

JI H++OH-

H2.5,

I HsSp + RsNH

I R2NH2+ + HS-

I
(Polysulfldes if 0s present) t

11
HS-

+HzO

I1 S*-

+ H+

reclaimer is omitted. The capacity of the average gas treatment unit is around 200000 Nm3 h-r of natural gas, which we have chosen as the design basis.
TABLE 1. Specification Capacity: Pressure: of the feed gases

200 000 Nm3 h-r 70 bar

yvzrn

co2 (vol.%)

The calculations are carried out for two gas compositions as shown in Table 1, a high H2S, high CO2 gas (I) and a low H2S, high COs gas (II). In the latter case, in particular, selective removal of HsS is very important. No other sour gases such as COS or HCN are assumed to be present. In all calculations H2S is removed to the pipeline specification of 4 ppm vol. [7]. No specification is set for C02. Aqueous solutions of either di-isopropanolamine (DIPA) or methyl-di-ethanolamine (MDEA) are used as solvents. Absorbers The absorption process is based on reversible acidbase reactions between H2S and/or COz. These reactions

Gas I Gas II Specification

7 5 4 x 10-4

7 10 None

enhance the rates of mass transfer of the acid components into the liquid phase and also provide absorption capacity. The reaction between HaS and the amine involves a proton transfer only and can be regarded as instantaneous and reversible [ 10, 141 : H2S + R,R?RsN + HS- + RrRaRaNH+ (1)

Owing to the forward and reverse reactions which are fast compared with the rate of mass transfer, the equilibrium (1) is established everywhere in the liquid. On the other hand, the reaction between CO* and an aqueous amine proceeds at a finite rate. The COz absorption rate is determined by the reaction [9,10,14] CO2 + 2R1RzNH for a secondary + R,R,NCOO+ R,ReNHa+ (2)

liquid phases) reactors, each corresponding to an actual tray (see Fig. 2). Process conditions and gas and liquid compositions are assumed to be uniform in each reactor. More sophisticated tray models, e.g. with plug flow in the gas phase as proposed. but not used, by Cornelissen [8] are not considered. Gas and liquid phase backmixing between trays as well as pressure drop over the trays are neglected. The calculation starts with an overall mass balance over the column. For this purpose a CO* concentration in the treated gas, J$&,, has to be estimated and then the treated gas flow, @+l, can be calculated by
,g+1= --

o:u- Yir,s- Ybo,

1 -YF+Y~~~

Nm3s-l

(5)

amine, and [9, lo] by = HCOs- + RIR2RSNH+ (3)

COz + Hz0 + R,R2RaN

a base catalysis mechanism for a tertiary amine. At longer residence times, the carbamate, RIRzNCO@-, formed from COa and secondary amines, is hydrolysed [ 141 and produces bicarbonate and free amine: R,R,NCOO+ Hz0 = RIRzNH + HCOa(4)

whereas the H2S concentration in the treated gas is always taken as 4 ppm vol. The composition and the flow of the treated gas is fixed now. Next, the liquid volumetric flowrate in the column is calculated in such way that the rich solution leaving the column at tray 1 is loaded with acid gases to a preset desired value atot:

%ot = aH,S

WzStot]Z, +
+ %O,
=

[co,~,,,]~ --

(6)

The rate of this last reaction, however, is low and therefore it is generally assumed that it does not affect the CO2 absorption rate significantly. Two types of high pressure absorbers will be compared: a conventional tray column and a cascade of trickle bed columns. The reason for incorporation of the latter absorber type will be demonstrated later. A tray absorber is used in the TGU calculations.

[Amtot

W%otI~~ [C%totl; and [Amtot] refer to the total concentrations of the respective components in both reacted and unreacted form. The liquid flow is obtained from a total acid balance:
& = {++([H2S];+ + + [CO,]:)

4&[HA: + W&)Y{%t
[H2S];+ ~ [C02];+}

[Amtot
m3 s-r (7)

Tray absorbers
The tray absorbers are calculated by a tray-to-tray procedure. The absorption column is considered as a series of ideally mixed (with respect to both gas and
treated gas t Lean solution

where all gas phase concentrations are expressed in mol Nme3, under standard conditions. We assume $Q to be constant over the column. From the overall mass balances for the individual components the total acid gas concentrations in the rich solution are calculated:

+ D-L%,,1~+
and 4[CO21: KO2,totlb = GQ + P332, totlPn+l -4;+W2Sl,+

mol me3

(8)

mol m-

(9) are of a heat overall that (10)

The heat capacities of the gas and the liquid the same order of magnitude and therefore balance has to be incorporated in the model. The heat balance over the absorber is solved assuming T+r = n+1 TQ s K

feed gas

I I

rich solution

Fig. 2. Scheme for the tray absorber.

which means that the temperature of the top tray is that of the lean amine solution introduced. This enables us to calculate the temperature of the rich solution, provided that gas and liquid inlet temperatures are specified:

6 T; = 5:;-- - 4; +T,+hgCpg ~__

&PQ&Q

q$[H& and

- @: tH~Sl:=J~,dtrw

mol s-l

(13)

~~~~,W~~,,,lb ____ +pQC,Q

PG%,,1P+> ---

6: [CO,]:

- 4: P-M:

= Jco,bw

mol s-l

(14)

+T;+

(11)

The AH terms represent the sum of the heats of both reaction and adsorption. The gas and liquid compositions and conditions at the absorber bottom are now fixed and provide the starting point for tray-to-tray calculations. In order to solve the tray heat balance, thermal equilibrium is assumed between the gas and liquid flows leaving the tray :
Ti+l= I

Using mass and heat balances for tray 1, [HZ&,,]& [CO *, t,t]i and Tz are then calculated. The tray-to-tray procedure continues until the H2S specification is met (4 ppm vol.). If the CO2 concentration in the treated gas deviates by more than 5% from the initially estimated value, the column calculation is repeated right from the beginning using the calculated COZ concentration as a new estimate.

Cascade of trickle bed absorbers

The second absorber type evaluated in this study is a cascade of trickle bed reactors (see Fig. 3). In the reactors, gas and liquid flow cocurrently downward over an inert packing. In this mode of operation flooding does not occur and therefore higher gas throughputs per unit cross-sectional area can be realized than in countercurrent operations. In cocurrent flow ultimately only one equilibrium stage can be attained in each reactor. This implies that for our very deep H2S removal several reactor beds are needed which have to be connected to provide overall countercurrent gas and liquid flows, as shown in Fig. 3. This involves the installation of additional pumps. In the trickle bed calculation procedure each bed is regarded as a series of ideally mixed reactor sections.

(12)

At the absorber top, this equation is contradictory to the assumption for the overall heat balance expressed in eqn. (10). This discrepancy is, however, negligible later. because T[+ E T,$,as will be demonstrated Now, as the gas temperature on the bottom tray 1 is known from eqn. (12), the liquid phase equilibrium constants, Henrys coefficients, and gas and liquid phase diffusivities are calculated. The liquid phase composition, more specifically the concentrations of unreacted H,S, CO? and amine, is obtained by means of the equilibrium model described by Blauwhoff and van Swaaij [9, 111. For DIPA, however, the equilibrium of the carbamate hydrolysis reaction (4) will not be (fully) established in the absorber owing to the low rate of this reaction. Therefore, the equilibrium model was modified for DIPA, in order to account for this effect. Next, H2S and COZ gas phase concentrations at tray 1, ie. in the gas leaving tray 1, are estimated. Subsequently, gas phase fugacities are calculated by means of the Soave-Redlich-Kwong equation of state [15]. The absorption driving forces are now known. The molar fluxes JHzS and Jco, are calculated either using analytical solutions of the mass transfer models of Secor and Beutler [ 161 for H2S and Hikita and Asai [ 171 for COz, which do not account for interaction between the liquid phase reactions, or using the numerical solution method for simultaneous mass transfer and interactive reversible reactions, developed by Cornelisse et al. [13]. It must be noted that the film model solution of Secor and Beutler [ 161, used for the H2S absorption rate, was modified to a penetration theory model solution according to the suggestion of Olander [ 181. Details are given elsewhere [9]. After the first estimation of the gas phase concentrations and the subsequent molar flux calculations, iteration proceeds using a Newton-Raphson technique, until the following implicit flux balances are satisfied simultaneously:

feed gas1

reactor

reactor

1 Pig.3. Scheme for the trickle

treated gas

lean solution

reactor

bed cascade

reactor.

7 Like in the tray-to-tray calculations, backmixing in the gas and liquid phases as well as pressure drop over the sections are neglected. By means of mass and heat balances over the cascade of reactors, which are essentially identical with those for the tray absorber, the temperature of the rich solution leaving the reactor 1 and the total HsS and CO? concentrations in it are derived. As far as the reactor balances for HsS and CO2 are concerned, the procedure now proceeds slightly differently. Flowing cocurrently in a reactor, the HsS gas concentration will decrease and the (unreacted) HsS liquid concentration will increase owing to the mass transfer. Ultimately, the mass transfer driving force of HsS will be zero and only COs will still be absorbed at a relatively high rate. In order to avoid this (for the selectivity) extremely unfavourable situation, calculations are stopped at a preset positive absorption driving force for HsS. This is realized, for example, for reactor i, by calculating the unreacted HsS concentration [HsS]b n+l, m the liquid from temperature, total HsS and COa concentrations using the equilibrium model mentioned earlier [9, 111. The HsS gas phase concentration, [HsS]i+, at which the reactor calculations are stopped, has been using an arbitrary but effective algorithm :
i,n+l

temperature. Under the prevailing conditions, the absorption reactions (1) and (2) are reversed, the acid gases are entrained with the water vapour and pass overhead where the water vapour is condensed and returned to the regenerator. The basis of the regenerator design used in this study is the method introduced by Ouwerkerk [7]. Two operating regimes for the regenerator are distinguished: heat limited and stripping limited operation. In the heat limited regime, the steam flowrate for the regeneration of the rich solution is entirely determined by the amount of heat required to provide the latent heat for the desorption reactions, the sensible heat difference between entering and leaving liquid flows and the reflux flow. The steam rate is described by [7]

f4 =hdfst~%i,s[A%tl ,t

AHH,s

+ %o,[AmtotlA&o2 + + PQC,OIP~ - TkHlfWt

+ &?flUX
where $J reflux
= R%t [~totl~Q&

kg s-r

07)

kg s-l by off-gas off-gas

(18)

R represents
R=-__

a reflux factor defined

moles of steam in regenerator moles of acid gas in regenerator

[H2S]; where

n+i = f dl:

a 1

(19)

mol Nmm3

(15)

[H~SIQ -mHas = [H,S],

A value of R = 1.2 is used throughout this study [ 11. Since the heats of reaction AHH*S and AHco, are almost equal [ 11, eqns. (17) and (18) may be combined to give

at the interface

(16) $2 = @dGt/ol,,t Wmtotl( 2 + !f?&%(TzUt_ Tp)

st In the stripping limited regime, regenerator operation is controlled by the minimum gas (steam) flow required to dilute the gas components at the regenerator bottom sufficiently to provide a driving force for the desorption. This minimum gas flow can be obtained from the operating line tangent to the equilibrium curve at the desired residual acid gas load or, (see Fig. 4) [7] : kg s--l (21) dW2$. I a, The operating line can be considered as a straight line near the regenerator bottom only where the heat needed for the desorption reactions, and hence the condensation of steam, is still negligible. At higher trays large amounts of heat are transferred to the solvent by condensation of the steam and consequently the operating line will curve upward. The solvent temperature at the regenerator bottom is assumed to be the boiling temperature of water at the prevailing pressure. Equilibrium curves are generated by the equilibrium model [9, 1 I] at this temperature, assuming equal amounts of H2S and CO2 in the lean solution. The higher value of either the heat limited or the stripping limited steam rate has to be applied.

+R)
kg s- (20)

and where f is a constant >l which may have different values for each reactor. Condition (15) ensures a positive driving force for the HsS absorption in each reactor section and terminates the reactor calculations when the H2S equilibrium is approached too closely. An initial estimation of the Cq, gas phase concentration leaving reactor i, [CO,]; +l, now provides sufficient data to complete the reactor mass and heat balances and the inlet data can be obtained. Next, a section-to-section calculation proceeds until the H2S gas phase concentration fails below the value calculated by eqn. (15). If the calculated CO2 gas concentration deviates by more than 3% from the initial estimate the procedure is repeated using the calculated value as a new estimation. New reactors are added until the HsS specification is met (Table 1). If the CO2 concentration in the treated gas deviates by more than 5% from the initial estimation, the overall calculations are repeated, It will be understood that the choice of the factor f and the number of beds and their lengths in this cascaded reactor are interrelated. This requires a suboptimization of the economics of this reactor type not presented here.

@,t=@Q&t

dffqs

[htotl

Regenerator
In the regenerator the H2S and CO2 components are stripped out of the solution at low pressure and high

mixture obtained, for example, of natural gas [ 1,2] :

by partial

combustion

SOs + CO t 2Hs =+ HsS t CO* t Hz0

(24)

Fig. 4. Determination limited regime.

of minimum

steam rate in the stripping

The volume of the catalyst in the reduction reactor is estimated by the same procedure as used for the Claus converters. The gas leaving the reactor is then cooled to 155 C in a waste heat boiler in which steam is generated. After further cooling in a feed water preheater the excess water vapour in the gas is condensed in a packed cooling tower by means of circulating water. The dimensions of the tower are determined by the procedure described by Treybal [21]. The cooled gas now enters the SCOT absorber, specifications of which are calculated by means of the procedure described earlier. The HaS specification for the absorber off-gas is set at 160 ppm vol. [22].

Sulfur recovery unit

The hydrogen sulfide in the regenerator off-gas is converted to elemental sulfur in a split-stream threestage Claus unit (see Fig. 1). In this configuration onethird of the Claus feed gas is fed to a burner and converted with oxygen from air into SOs according to the exothermic reaction [ 191 HaS + (3/2)0, =+Ha0 + SOi (AH= -5 19 kJ mol-)

Heat exchange equipment

Parameters of the heat exchange equipment are not calculated in detail. Overall exchange areas are obtained by the e-NTU method described by Kays and London [23] for countercurrent operations. Heat transfer coefficients of 600 W me2 C- are used for liquid-liquid heat exchange, of 1500 W rn- C-l for exchangers involving condensing or evaporating media and of 300 W rn- C for condensing vapours containing non-condensable gases.

(22) The larger part of this hot gas stream is cooled in a waste heat boiler to such an extent that after mixing with the unconverted main gas stream the temperature of the gas to the first catalytic converter is 230 C, the optimum inlet temperature for the first converter [2]. A small part of the hot gas stream is bypassed to reheat the feed to the second and third converters (see Fig. 1). In the catalytic converters elemental sulfur is produced by the equilibrium reaction [ 191 2Hs.S + SOa == 2HaO + (3/x)$ (AH= -147 kJ mol-)

Physico-chemical data
A large amount of physico-chemical the absorber calculation procedures. sources of data and pressure/temperature will be summarized briefly. data is used in In this section dependences

The liquid phase

The composition of the liquid phase at each tray or section is calculated by the equilibrium model of Blauwhoff and van Swaaij [ 111. Equilibrium constants and Henrys coefficients for this model as a function of temperature are obtained from Edwards et al. [24], Blauwhoff and Bos [ 121, Schwabe et al. [25] and Blauwhoff [9]. Pressure corrections for Henrys coefficients are taken from Edwards er al. [24]. Rate constants for the C02-amine reactions are obtained from Danckwerts and Sharma [I417 Blauwhoff et al. [lo] and Blauwhoff [9]. Heats of reaction are summarized by Kohl and Riesenfeld [ 11. The liquid phase viscosities for aqueous DIPA and MDEA solutions were measured by Blauwhoff [9]. The simple Andrade correlation is used for extrapolation to higher temperatures [26]. Pressure dependence of the viscosity is considered to be negligible in the relevant pressure range [26]. The diffusivities of HsS, CO2 and DlPA in aqueous solutions are correlated with the liquid viscosities and temperatures by amodified Stokes-Einstein relation [26] :

(23) For reasons of simplicity, we assumed S, = Ss, which is a good approximation at T < 330 C [ 191. The conversion of H$ in each catalytic converter is assumed to be 95% of the thermodynamic equilibrium conversion. After each converter the gas is cooled to 155 C to condense the sulfur vapour. The removal of sulfur from the gas stream favours conversion of HaS by the equilibrium reaction (23) in the next converter, which is entered after reheating the gas to 230 C by the bypassed hot gas. The dimensions of the converters are derived by the relation for the required amount of catalyst either by Kohl and Riesenfeld [ 1] or by Fisher [20]. The most conservative of the two results is used in our calculations.

Tail-gas unit
In the tail-gas unit, based on the Shell Claus off-gas treatment (SCOT) process, SOa in the gas leaving the SRU is first catalytically reduced to HsS [l]. This reaction occurs at -285 C by means of a CO/H? gas

D7)23 __ = constant T

(25)

9 The diffusivities in pure water serve as a basis for this correlation [14,27]. The diffusivities of the ionic species in the solution and of MDEA are assumed to be equal to the diffusivity of DIPA [14]. Properties not mentioned explicitly are taken to be the same as for water under the prevailing conditions.

k, = 8.0 X 1O2 $-o*4 Resoe4 Scs.

m s-l Since we considered the influence (30)

The gas phase

The gas phase is described by the Soave modification of the Redlich-Kwong equation of state [ 151. Constants and critical properties for this equation can be found in Reid et al. [26]. Gas phase diffusivities are calculated by the Chapman-Enskog relation [26] and are assumed to be inversely proportional to the operating pressure.

The mass transfer parameters

High pressure mass transfer coefficient and interfacial area correlations for tray and trickle bed absorbers are not available in the open literature. Therefore we used correlations for the dependence of the mass transfer coefficients on the hydrodynamic conditions, derived at and extrapolated to higher atmospheric pressures, pressures. For the tray absorber calculations we employed the sieve tray correlations of Sharma and Gupta [28] as a basis for the influence of vp, vs, pp, pa and Heco* on kpa and k,a:

of the ratio d,/d,on high, we replaced this ratio by a fixed value of 0.075 in our calculations. This is a typical value for the geometries used by Fukushima and Kusaka [30-321. For the interfacial area we have chosen a conservative value of 200 m2 per m3 of reactor for all trickle bed calculations. This value corresponds to the geometric packing area of 1 inch Raschig rings. The mass transfer coefficients are converted to the operating pressures and temperatures by assuming a penetration theory dependence of kp and k8 on the diffusivity:

k* and k, to be unrealistically

m s-l

(31)

where D* refers to the diffusion coefficient under the conditions at which the mass transfer coefficient relations (26)-(30) were obtained. The interfacial areas were assumed to be independent of pressure [33] and temperature.

kQa = ~~(v~p~)~%,~~*
and

s-1

(26)

k,a = cz(vapa)06v,-2 Heco,

SC

(27)

Results of the High Pressure Tray Absorber Calculations Comparison of the non-interactive and interactive mass transfer models
A set of calculations for the high pressure tray absorber model was carried out in order to compare the results for the analytical non-interactive and numerical interactive solutions of the mass transfer model. In the analytical solution the HaS and CO2 liquid phase reactions are regarded as non-interacting and hence the mass transfer rates of H2S and CO2 are independent of each other. The numerical solution method, however, does account for interacting reactions. Calculations were carried out using feed gas composition I (see Table 1) and two amine solutions of 2.0 M DIPA and 2.0 M MDEA at absorber conditions as specified in Table 2, in order to compare both methods. The tray-to-tray gas phase compositions calculated by both methods are shown in Figs. 5 and 6 for DIPA and MDEA respectively. The difference between the results obtained by the two methods is only marginal. The maximum deviation of 13% in the COa concentration in the treated gas is found when using DIPA. A comparison of the two methods of calculation can be illustrated more clearly on the basis of the overall mass transfer resistances for H2S and CO*, as defined by

where cr = 4 9 X lo- s*s k g -o.~ and ca = 1 .O X lom3 bar sa.a kg-O:6 mol- at 70 bar and cr= 5.5 X lo-* ss kg-Oa6 and c2 = 1 .I X lo- m3 bar so-a kg.6 mol- at low pressures respectively. The gas-liquid interfacial area per m2 tray is calculated with the correlation of Nonhebel [29] for sieve trays:

a = 30(p,v,)0*5p,0*a5

m2 (m2 tray)-

(28)

By combination of eqns. (26) (27) and (28) and assuming an average froth height of 0.3 m, values of the mass transfer coefficients kp and k, are obtained. We are aware that using different sources for the derivation of kp and k, may lead to erroneous values for these parameters. In fact, we adjusted the values of ci and c2 in eqns. (26) and (27), as compared with the original values [28], in order to obtain realistic values for k, and kp in the relevant pressure ranges while maintaining their dependence on the hydrodynamic conditions. In the calculation procedure for the cascade of trickle bed reactors the correlations of Fukushima and Kusaka [30-321 for kp and kg in the pulsing flow regime are incorporated:

kp = 5 X lo2 $* Ree*33 Re,.

SCQ.

RTV=Ri+Rf=
m s-l and (29)

[Gilg - ICiln/W Ji
1 c-z__ k 0 1 k,, i + 1 s m-l (32)

m&Q, iEi

10
TABLE 2. Standard design parameters for the high pressure tray absorber General P and operating conditions

[A%otl Gasin : 4 =:
Gas out: Liquid in: DIPA: MDEA: Y;;,:Q n+1 Tp WzStotl;+

70 bar 2 kmol mm3 55.56 Nm3 s-l (= 200 000 Nm3 h-t) 293 K 4 ppm vol. 313 K 20 mol 20 mol 10 mol 10 mol 0.7 Fig. 6. Comparison of tray-to-tray gas phase compositions for the interactive and non-interactive mass transfer models (MDEA, feed gas 1).

KO?.tOtl;+ IWtotl~+i Koz.tOtla+ C&

me3 me3 rnp3 rnp3

Liquid out :

in which

Icil II mi= [cilg Iatinterface

(16)

Et is the enhancement factor obtained from the analytical and numerical mass transfer model respectively. The calculated overall resistances are shown in Figs. 7 for DIPA and 8 for MDEA, together with the temperature profile. On the first few trays in the bottom section, both the H,S and COa driving forces for the absorption are high. On the other hand the solution is loaded with acid gases, so that the concentration of unreacted amine, the alkaline liquid phase reactant, is low. These two factors cause the amine transport from the liquid bulk into the reaction zone near thegas-liquid interface to be controlled by diffusion, so that the amine is depleted in the reaction zone. Owing to this depletion HaS has to diffuse into the liquid phase before it can react with amine. Consequently, the H2S transport 10-l analytical sdutm
metkd

330

320

Tl

3iO

LKI

J
Pig. 7. Comparison of overall interactive and non-interactive feed gas I). mass transfer resistances mass transfer models for the (DIPA,

OKWJ,

11

8 -trZly-

12

16

20

24

Pig. 5. Comparison of tray-to-tray the interactive and non-interactive feed gas I).

gas phase compositions for mass transfer models (DIPA,

is partially liquid phase limited at the bottom trays both for DIPA and MDEA. Higher up in the column, the H2S gas phase concentration along with its driving force falls rapidly (see Figs. 5 and 6). The unreacted amine concentration increases upwards to the absorber top. This results in an increasing H2S enhancement factor EHIS, so that the overall mass transfer resistance Rfi,s drops according to eqn. (32). Finally, at very large values of EH s, the overall mass transfer resistance approaches the gmit of

11 CO?--amine rate of reaction. Obviously this is levelled out in the results obtained by the numerical method as caused by the HsS co-absorption. At first sight, it is surprising that the analytical mass transfer methods of calculation, which do not account for the interaction of the liquid phase reactions, agree so well with the more sophisticated numerical method. This can be understood, however, by realizing that at the column bottom the molar fluxes of H2S and COa are high, so that the mass transfer rates are determined only by the diffusion limitation of the amine and therefore are independent of the rates of the interacting reactions. In the upper part of the column the HIS mass transfer rate is exclusively gas phase limited and, therefore, independent of the model chosen for the liquid phase, whereas the CO* absorption rate is not influenced at all by the simultaneous H2S absorption due to the low H2S molar fluxes. Because of the good agreement between the two methods, wherever possible, we thenceforward used the analytical method which consumed the least computer time. Influence of design absorber performance and operating variables on the

(
)

.-.,
i

I..

k.
1.y
k,.
... -..._.

30

20

xl-

--_ -. -

.I ---.- ...._._.___._,,,__~_,__ Tl

I Tl WI

IO

----__

---------

a852

M-

DC-

xl-

00.

oo-

IO-

\\ \\ -___ ___--- _ 6%

of+

\-;Rs
1

c
-hay

12 -

16

20

24

Pig. 8. Comparison of overall interactive and non-interactive feed gas I).

mass transfer resistances for the mass transfer models (MDEA,

Extensive calculations on the high pressure tray absorber were carried out in order to determine the influence of a number of design and operating parameters on the overall absorber performance. Two parameters are used to characterize this performance: firstly, the overall absorber selectivity defined by moles HsS absorbed 17=

complete gas phase resistance indicated by the lines drawn in Figs. 7 and 8. Generally speaking, the H2S mass transfer in the high pressure absorber can be considered to be gas phase controlled on a column scale. Both model calculations agree well for HsS in the regime where distinctions-if present-would have been expected, i.e. in the partially liquid phase limited mass transfer regime at the absorber bottom. The HsS mass transfer is obviously entirely determined by the diffusion rate of amine and therefore independent of the model used. In the gas phase limited regime (tray > 8) of course, no distinction between the models is observed. Only at the top of the column the H2S overall mass transfer resistance calculated by the numerical method slightly increases. This effect is due to a decrease of the H2S enhancement factor, caused by the relatively high COZ molar flux. For the CO* mass transfer resistance profiles the situation is different. In general the COs abso tion T rate is almost exclusively liquid phase limited (Rco, > 0.8R$bZ for DIPA and R&-,*> 0.9R& for MDEA). As for H2S, the two methods of solution agree well and only minor distinctions can be observed. At the bottom tray the amine depletion is extreme and therefore the COs overall mass transfer resistances are highest here. At higher trays the depletion decreases and the resistances fall steadily. With the analytical method a local minimum value of Rz& is observed at tray 2. This is caused by the high liquid phase temperatures (see Figs. 7 and 8) and the correspondingly high

moles H2S + COZ absorbed

and, secondly, the solvent flowrate &. The first parameter r~ equals in fact the fraction H$S in the regenerator off-gases and therefore virtually controls the dimensions of the process equipment linked to the regenerator in the SRU and the TGU. Low values of q result in low H2S and high inert COZ concentrations in the SRU feed gas and have a detrimental effect on the dimensions of SRU and TGU. The minimum value of n is obtained when all acid gases are totally absorbed and is solely determined by the ARU feed gas composition: (34) The values for nmin are 0.50 and 0.333 for gas compositions 1 and 11 respectively (see Table 1). The maximum value of the selectivity on the other hand is unity, if no COs is absorbed at all. The lower the selectivity the more COs is co-absorbed and the more steam is wasted for the CO1 desorption in the regenerator. Thus the incentive for improving the absorption process increases rapidly with decreasing n.

12 The second parameter $Q at a constant acid gas loading ctztot controls directly the amount of steam required for the regeneration of the rich solution, as can be seen from the eqns. (20) and (21). As steam costs are the major item in the operating costs for gas treatment, we must reduce the solvent rate as far as possible. In the Figures to this section these two parameters 9 and & together with CO2 concentration in the treated gas and the number of trays required for the H2S specification, are plotted as a function of the most important design and operating parameters. The design conditions are summarized in Table 2. Operating pressure of the tray absorber The operating pressure of the tray absorber affects a number of parameters. By increasing the pressure the driving forces for the H2S and CO* absorption become approximately proportionally larger, whereas the mass transfer coefficients and interfacial areas decrease owing to the lower volumetric gas throughput (see eqns. (26)(28)). Moreover, the gas phase diffusivities decrease and hence the gas phase mass transfer coefficients will also decrease (see eqn. (31)). The overall influence of the operating pressure on the absorber performance is shown in Fig. 9 for the feed gas compositions I and II and for DIPA and MDEA solutions. In general it may be concluded that the pressure does not affect the absorber performance. For the DIPA solutions the selectivities for the gas compositions I and II are close to the corresponding minimum values 0.50 for gas I and 0.33 for gas II, as is also indicated by the low COz concentration in the treated gas. By using MDEA solutions the absorber operation is only marginally more selective (1 l%-14%) and the solvent flowrate spondingly by reduced. at constant atot can corre-

The regenerator effectiveness-the the lean solvent, a& For reasons of simplicity, in the regenerator bottom of H2S and COz:

acid gas loading of

we assumed the lean solvent to contain equal quantities

&tot = (yH,S + %O,

where QI,s = ace, or

(35)

b-k&tl~ = [COz,t,Ja

(36)

The results of the calculations at varying acid gas loadings of the lean solvent are summarized in Fig. 10.

0 OIW, gas I .OIPA,garII---

xM)EA,garI

-.-

+mA.

gas II

..--

_..-..-.._-

_-..-..-

G;
(9/.)

2-O-

_.-._.-.-.-.-.-_-.-

-.-.-.---.-. I---------------A...

qmm.

gas,
II

..__ .__.

-.-.-

.._. -._,,__

_.._.---=2~~~.~=--..

lmmgas

io

26

i0

IO-------________

lH2S,totlln;lCO2,t0tlIn~ ImolelmS

Fig. 10. The influence of the acid gas loading of lean solvent on

027

------1 _______
..__ __._______.___._.________ _..-. _ ._.._..-..--------------.
~min,gasI

tray absorber performance. An increase of the acid gas loading is detrimental to the driving force for H2S absorption at the top of the column, whereas the CO* driving force is hardly affected. To meet the H2S specification more trays are needed and therefore the CO* co-absorption also increases. As a result the selectivity decreases and the solvent flow increases. For DIPA solutions these effects are less marked since the absorbers already operate close to minimum selectivity and maximum solvent flow. For MDEA, however, the effects are quite pronounced and these solutions, therefore, should generally be regenerated to lower acid gas loadings. The optimum atot follows from an optimization of the steam consumption in the regenerator. This latter subject will be discussed later.

0,

04

_._ _ _._._.-.-.-.-.-.i----------------------------.._.-..-

so

60

70

80

P brl

Fig. 9. The influence absorber performance.

of the operating

pressure on the tray

13

Acid gas loading of the rich solution, cxiOt

The acid gas loading of the rich solution is a very important parameter in the absorber operation. Results are shown in Fig. Il. An increase of the e&t results in a reduction of solvent rate $Q without affecting the process selectivity. The approximately constant CO* concentration in the treated gas, r,$&, implies that the total acid gas pick-up by the solvents is constant at varying riot. If the acid balance over the absorber (7) is rewritten in the form de]Amt,tl(& = - 4.Z)

TABLE 3. Values of the constants

in kmol s-t

in eqn. (38)

Solvent DIFA Gas I Gas II 0.352 0.314 MDEA 0.315 0.326

G:UWl; + [CW:) - $;+{[HsS],n+ + [CO&+}

mol s-l

(37)

and as the right-hand side of this equation appears to be constant, relation (37) can be simplified to #a]Am,,,](Q tot ~--&,;) = @Q[htotl = constant
a:ot

(38)

This equation relates (Y& directly to the solvent rate. The constant in eqn. (38) is obtained from the results of the absorber calculations and is summarized in Table 3 for all four gas-solvent combinations. In heat limited regenerator operation, the reduction of steam consumption by increasing Q:,,~ will generally be rather low. This is caused by the fact that, although the solvent flow is reduced, the product 4Qatot is fixed (see eqn. (38)) and hence the amount of steam required to provide for the heats of reaction and the reflux is
00IPA.gas I l wn.4as II--ways XP-UEA.~~SI - -.+MoEApS II---- .-

YCY; I%1

20 ,.

.- -.- _ -.-.-.._ .-.._.._.,_ 1

-.-.-.-.-._._._._._._ _._ i -----------------

also almost fixed (see eqn. (20)). In this heat limited regime of the regenerator operation, the reduction of steam consumption will therefore only originate from the smaller amount of sensible heat to be supplied to the solvent. In the stripping limited regime the steam consumption is proportional to the solvent flowrate only (see eqn. (21)) and hence inversely proportional to c& at constant [Amto,]. An increase of &, in this regime is far more effective in reducing the steam consumption than it is in the heat limited regime. Though not shown explicitly in Fig. 11, there exists a maximum limit of the acid gas loading for given operating conditions. This maximum is determined by operation of the bottom trays of the absorber. With increasing gas loading (Y&, the liquid phase temperature at the bottom trays rapidly rises. The combined result of increased loading and rising temperatures causes the unreacted HsS concentration to be higher and reduces the H2S absorption driving force. The maximum acid gas loading is obtained if the driving force at one of the bottom trays becomes zero and a pinch between equilibrium and operating hnes of the absorber results. This phenomenon is not bound to occur at the first tray, because its temperature is usually appreciably lower, owing to the introduction of the cold feed gas, than the temperatures of 2nd and 3rd trays (see Figs. 7 and 8). For our specific calculations the maximum acid gas loading will be around OL{,~~0.85.

Total amme concentratron

[Amto,

Fig. 11. The influence of the acid gas loading tray absorber performance.

of rich solvent on

The total amine concentration is a parameter with a most complicated influence on the absorber operation. In Fig. 12 results are shown for varying amine concentrations. The influence ofthe amine concentration is roughly twofold. Firstly, increasing [Am,,,] proportionally decreases the solvent flow, identically to the effect of the acid gas loading of the rich solution. Since the COs concentration in the treated gas is almost constant, eqn. (38) and the constants summarized in Table 3 can be used for correlating the solvent flow and the amine concentration. With respect to the reduction of the steam rate in the regenerator as affected by lower solvent rates the same remarks as were made for the influence of criot are valid. Secondly, with increasing amine concentration the solvent flow and therewith its capacity for heat absorption are reduced; consequently, the temperature profdes over the column become more pronounced. This is clearly demonstrated in Fig. 13 for three MDEA solutions and for gas I. With increasing MDEA concentration the absorber temperature level

W2Slg

and

M-

MzSl(mtt2s lmolelm? 20 G; 1%) 1

1.0

1.

_._. _. - _._

_ -.-.-.-

Ol-

1 -------------

0 01.

o.oolJ .

s
-tray

12

16

20

24

Fig. 14. HsS concentrations in the gas phase and unreacted in the liquid phase as a function of the tray number and the MDEA concentration (oFztt = 0.7).

particularly

2-o

25 lamtot

3-o

3-s Ikmolelm?

Fig. 12. The influence absorber performance.

of the amine

concentration

on tray

3704

in the absorber bottom, are reduced. Consequently, the HaS molar fluxes are reduced and additional trays are required to meet the H2S specitkation. The CO2 concentration profiles show a more significant sensitivity towards the temperature or MDEA concentration (see Fig. 15). As a result the COz liquid concentration, [CO,] a/mco,, rapidly increases with the MDEA molarity in the lower part of the absorber. For the 3.2 M MDEA solution this concentration even exceeds the gas phase concentration at trays 2, 3 and 4, resulting in a desorption of COa and increasing CO? gas concentrations. Unfortunately, this selectivity favouring effect is almost completely nullified by the larger CO* driving forces at higher trays and the extra trays required for the H2S specification. Similar results are obtained for DIPA solutions.
zoo-

ICO21g

a IlPICQ-

-tray

~~02hhca2

Fig. 13. The influence of MDEA concentration profiles in the tray absorber.

on temperature

imoleb? 80-

rises and the maximum shifts to higher trays. This change of temperature profiles affects the liquid phase equilibria and thus the absorption driving forces. In Fig. 14, gas phase and unreacted liquid phase concentration [H2S], and [HzS]p/mnIa, respectively, are plotted as a function of the tray number for the MDEA concentrations. The difference between the gas and liquid phase profiles represents the HIS absorption driving force. Be aware of the logarithmic scale for the concentration. With increasing MDEA concentration [H,S]a/ mHzS increases and the absorption driving forces,

60.

4O-

zo-

-tray

Fig. 15. CO2 concentrations in the gas phase and unreacted in the liquid phase as a function of the ?ray number and the MDEA concentration (ot, = 0.7).

15 With an increase of the amine concentration at a futed acid gas loading oiOt, unreacted HaS and COa liquid phase concentrations rise rapidly and reduce, in particular, the absorption driving forces at the absorber bottom, as has been shown already for MDEA (see Figs. 14 and 15). The increase of amine concentrations is limited to the point where the absorption driving force for H2S at the absorber bottom is reduced to zero and where a pinch between equilibrium and operating lines is developed. This limit is attained for c& = 0.7 at -3 M DIPA and -3.5 M MDEA molarities both for gas I and II. bar. From the aspect of this k,/ka ratio, tray absorbers perform worse than almost any other type of gas-liquid reactor. Therefore we compared the tray absorber with a cascade of trickle bed reactors. The k&a ratio for this type of absorber is typically around 80 at atmospheric pressure and around 10 at 70 bar and leaves room for further optimization owing to the absence of flooding phenomena. This improvement in the kg/k* ratio, however, may be partially offset by the fact that in each reactor HeS equilibrium is approached, which is undesirable from the point of view of selectivity.

General Remarks
From the results discussed it is evident that the three parameters I$,:, (Y&,~ and [Am,,,] affect the solvent flowrate and only orOT influences the selectivity. Although not every reduction of the solvent rate results in a proportional decrease of the steam rate, it is in general favourable to reduce the solvent rate as far as possible. This is realized by operating at the highest possible acid loadings in the rich solution and amine concentrations and at the lowest possible gas loadings in the lean solution. Whether the latter is feasible can be judged only on the basis of an overall evaluation of the economics. The, commonly referred to as selective, tertiary amine MDEA in general is some lo%-15% more selective and requires corresponding lower solvent rates than DIPA under our process conditions. High pressure tray absorbers using MDEA solutions still operate fairly close to the minimum theoretical selectivity. A closer look at our results shows that the enhancement factor for the CO1-aqueous MDEA absorption is rather close to unity-typically 1.1 < ECO, < 1.3. This implies that the COa mass transfer rate is controlled by the liquid phase mass transfer coefficient and the absorption driving force only and can be expressed by Jco, = kp P2lg -

Comparison between the Tray Absorber and a Cascade of Cocurrent Trickle Bed Reactors
The performances of the tray absorber and cocurrent trickle bed cascades are compared under the standard conditions given in Tables 2 and 4, respectively. The results of the calculations are summarized in Table 5 for two gas compositions and for both DIPA and MDEA solutions. In all situations the trickle bed cascades perform better-in terms of selectivity and solvent rates-than their tray absorber counterparts. The solvent rates can be reduced by around 60/o-8.5% for DIPA and around 9%-18% for MDEA solutions owing to the improved H2S selectivities. Moreover, the absorber diameters required are much smaller, which is an advantage in the plant economics.
TABLE 4. Standard design parameters for the trickle bed cascade reactor General: P [Amtot1 Gas in: @: T: Gas out: and operating conditions

70 bar 2 kmol m+ 55.56 Nmp3 s- (= 200 000 Nm3 h-l) 293 K 4 ppm vol.

[co2lQ
mco,

mol mm2 s-l

(39)

Liquid in DIPA: MDEA:

n+1 TQ [Hz%otlQ [CO2,totlQ

n+1 n+1

On the other hand, the H2S mass transfer rate is gas phase limited in the larger part of the absorber and is thus described by JH,S = kg [H,$g ___ mH,S

[%&IQ
[COP. Q

lZ+t

totli+1

[HA Q

313 K 20 mol me3 20 mol me3 10 mol rnp3 10 mol rnp3 0.1 m s-l 0.7

mol me2 s-l

(40)

Liquid out

dot

Since the absorber selectivity is directly related to the ratio of JH+ to Jco2, the selectivity for MDEA can obviously be improved by increasing the ratio of the H2S and COa driving forces and the ratio kg/kg. The first ratio may be improved by the application of different amines which almost do not react with C02, resulting in high [CO~]Q values. Further improvement in this respect may be obtained by the use of solvents which affect the solubilities mu+ and mco,. In this study we focused our attention on the second ratio, kg/kg. Using tray columns this ratio is typically around 50 at atmospheric pressure and around 6 at 70

The difference in the absorption characteristics between the tray and cocurrent trickle bed absorbers is illustrated on the basis of gas concentration and mass transfer resistance profiles which are plotted as a function of the cumulative interfacial area in Figs. 16, 17(a) and 17(b) respectively. Owing to higher mass transfer coefficients and hence lower mass transfer resistances the gas-liquid interfacial areas required to bring the gas on H2S specification in the trickle bed reactors are reduced to less than half of the area needed in the tray absorber (see Fig. 16). The higher k,/kQ ratios

16 TABLE 5. Comparison of dimensions and performance of tray absorbers Gas composition Amine
Reactor type Absorber diameter (m) Number of trays, sections Absorber length (ml Total interfacial area (m) Selectivity n SOlVentflow @IQ (m3 s-l) YEo,(%)
@Q, trickk/@Q, %ricklehray

I
DIPA Tray 3.45 22 19 6400 0.511 0.250 0.30 0.944 1.070 Trickle 1.75 5 17 3050 0.547 0.236 1.21 MDEA Tray 3.34 22 19 6120 0.568 0.225 1.68 0.911 1.121 Trickle 1.62 5 17.5 2890 0.637 0.205 3.01

II
DIPA Tray 3.55 22 19 6600 0.346 0.265 0.53 0.917 1.107 Trickle 1.75 6 19 3100 0.383 0.243 1.96 MDEA Tray 3.40 23 19.5 6.550 0.393 0.234 2.29 0.821 1.237 Trickle 1.56 6 21 4000 0.4 86 0.192 4.71

tray

kg/kg %

vQmoal*

or

E vg-o18

(42)

--.-.-

- --

DIFn,tray MOEP,hy wA,tnrkle-bed HlEA,trlchle-bed

According to eqn. (42), a superficial velocity which is as low as possible is required in order to obtain a high k,/kp ratio and hence a high selectivity. This inevitably leads to larger trickle bed diameters and hence to larger investments. An economic optimization is needed to determine the optimum design and operating conditions as far as the k,/kp ratio is concerned. In fact, vp and vg are only directly coupled in the first reactor where the feed gas enters and the rich solution leaves the cascade at a preset acid gas loading cx~~~(see Fig. 3). This implies that the effective superficial liquid velocities in the following reactors (2 to n) may be

4 o-ooo'o

s
2doo

12 4doO

1p

2p
6600

25

fray g-I area kn21 40

Fig. 16. Gas concentration profiles as a function of the cumulative interfacial area in tray and trickle bed absorbers (DIPA and MDEA, gas I).

7 .____------~___________

realized in the trickle beds result in an improvement of the ratio of mass transfer resistances Rg&/Rf&, which forms virtually the basis of the higher selectivity obtained in the cascade of trickle bed reactors (see Figs. 17(a) and (b)). The influence of the design and operating parameters p, &not, (Y:::, and [Amtot] on the trickle bed absorber performance is roughly identical to that for the tray absorber set-up. However, in the design of cascades of trickle bed absorbers additional degrees of freedom are available, i.e. the superficial gas and liquid velocities, the packing size and the equilibrium approach factor f used in eqn. (t5). The first three degrees of freedom can be used to optimize the k&a ratio even more. This ratio can be derived from eqns. (29) and (30) and predicts the following proportionality after elimination of d,: (41) Since in our set-up the total amounts of liquid and gas pass through all reactors, vp and vs are directly proportional to each other by the overall mass balance eqn. (7), so that virtually only one really effective parameter remains:

20

cwA.tray ---- MOEA.tray -.-.- DIR,tnchle-bed ..-..-MDEA.+nckle-bed

I
1
0

.._-
~_.__ _.._.._-..-.r-_._._.___.i-

62s

s
2dOo

12 k&O

16

20 Sob0

24

tw

g-la-a (rn2i

pig. 17(a). Overall mass transfer resistance profiles as a function of the cumulative interfacial area in tray and trickle bed absorbers (DIPA and MDEA, gas I).

17

_,.-.-_--15,/ -. \

._

--.- ..-..-

t-tOEA,tray
DIR. trckle-bed MOEA.trockle-bed

own., tray

10

_c_------______
/

,I :

s
x03

12 IdNo

16

20 6000

tray

0 10 20 1tmtot 30 1; =lco2,totl; 40 holeim31 54

0 0 g-t area lm*l

Fig. 17(b). Ratios of overall mass transfer resistances RE&,/Rg2~ as a function of the cumulative interfacial areas in tray and trickle bed absorbers (DIPA and MDEA, gas I).

Fig. 18. Steam consumption curves for the regeneration 2.0 M DIPA solution (c&t = 0.7,& = 0.25 m3 s-l).

of a rich

reduced. This will result in an increased k,/kQ ratio for these reactors and in a reduced CO* removal per reactor as well. Again an evaluation of the economics will yield the optimum superficial liquid velocities for each reactor. The last degree of freedom mentioned above, the equilibrium approach factor f, hardly affects the selectivity of the operation, but largely controls the layout of the cascade. The optimum values off for each reactor bed can be determined by an evaluation of the economics and will probably result in a set of beds of equal lengths.

Results and Discussion Calculations

of Regenerator

The regenerator operation is illustrated for the regeneration of the rich DIPA solution after absorption, as summarized in Table 5, column 1. Some results are shown in Fig. 18. In this Figure, two sets of lines can be distinguished: the set of horizontal lines (HL) describes the heat limited operation of the regenerator according to eqn. (17). It can be seen that the heat transfer in the lean/rich exchanger or the temperature difference between the leaving and entering liquid flows (temperature approach) has a substantial effect on the amount of heat limited steam required. The lower this temperature difference the lower the amount of steam required, but at the same time the investment in the lean/rich exchanger increases. The second set of lines (SL) represents the amount of stripping limited steam required as calculated from eqn. (21). As mentioned before, the amount of steam required for the regeneration of the rich solution is the higher of either the stripping or the heat limited steam rate. For example, at a regenerator pressure of 1.5 bar and with a temperature approach of 8 C, the regenerator operates in the heat limited regime, if we want to achieve residual acid gas loadings of around 22 mol m- and higher. However, if a steeper reduction of this residual acid gas

loading is desired, the steam rate is determined by the stripping limited curve at 1.5 bar. If, for example, a residual loading of around 15 mol mP3 is required, under the above conditions (1.5 bar and AT= 8 C) around 105 kg of steam per m3 of solvent is required. However, by increasing the pressure in the regenerator to around 2.5 bar, the operation obviously becomes heat limited again and only around 74 kg of steam per m3 of solvent is needed! Since the regenerator operates at the boiling temperature of water at the prevailing pressure, the higher temperature at 2.5 bar (401 K) as compared with that at 1.5 bar (385 K) provides a substantially increased shift in the equilibria of the desorption reactions (1) and (2). This shift produces more free, now unreacted HIS and CO2 in the solution and increases the desorption driving forces, even though the operating pressure is higher, ie. the temperature effect on the equilibria overrides the pressure effect on the gas phase concentrations. In all our following calculations we used this effect and increased the regenerator pressures in order to reduce the steam rate to the heat limited minimum.

Economical Treatment Introduction

Consequences

for Sour Natural Gas

In this section, the eight absorber-solvent-gas combinations, which have been compared with respect to their technical performance in Table 5, are evaluated with respect to the economics. This necessitates an analysis of the economics, not for the absorbers alone, but for the entire gas treatment plant including the absorber-regenerator unit (ARU), the sulfur recovery unit (SRU) and the tail-gas unit (TGU). The evaluation of the economics requires an estimate of the investments in process equipment and working

18 capital and the manufacturing costs per unit product treated. The above will be dealt with for a treatment plant to be situated in The Netherlands with an operating time efficiency (0~~1 of 90%. Estimate of investments in process equipment as proposed by Lang. The Lang factor obtained in this way overestimates the investment in special, nonconventionally costed equipment such as the high pressure absorbers in our study. For this type of equipment a modified approach is used. (a) Columns and (reactor) vessels In cost estimates of absorber/regenerator columns and (reactor) vessels the equipment is divided into the (pressure) vessel and the internals. The cost of the (pressure) vessel is generally correlated either with the vessel weight [37-391 or the vessel dimensions [35,40]. We used a weight based correlation, given by Peters and Timmerhaus [37]. The effective cross-sectional areas of the tray absorbers and regenerators in the flow scheme are obtained from gas and liquid loads using the flooding correlation given by Treybal [21] for a 0.75 m tray spacing. These effective areas are then multiplied by a factor of 1.25 for downcomers, etc., to obtain the actual areas. The column height follows from the number of trays, the 0.75 m tray spacing and a 3 m length added for gas and liquid inlets and outlets at the top and bottom of the column. The number of trays for the regenerator is set at 20 [ 11. The length of the cascade of trickle bed absorbers is decided from the number of beds required to reach the HsS specifications and the height of each bed, adding a length of 2 m per bed for inlets and outlets. The dimensions of the flash vessel follow from the residence time of the liquid, which is set at 3 minutes in this study. For the reactor vessels the dimensions are obtained from the respective catalyst inventories. The weight of the pressure vessels is determined using the ASME standard for the calculation of the wall thickness. The weight thus obtained is multiplied by a factor of 1.2 for manholes, flanges, etc. The installed costs for internals are added to the vessel costs. The expenditure on sieve trays as used in the absorbers and regenerators is averaged at US S 1600 per unit at 3 m diameter. Correction for different diameters is taken to be proportional to the column cross-sectional area. For the trickle bed absorbers packing costs are taken at US $ 900 per m3. An example calculation for the absorbers is given in Table 7. (b) Other items of equipment Heat exchange equipment costs are related to the exchange areas [35]. Correction factors are incorporated

To estimate the investments in chemical process equipment and plants several methods are available in the literature. The degree of accuracy of these methods depends by and large on the amount and reliability of the information on plant design. Three categories of methods of estimation may be distinguished: _ order-of-magnitude estimates (typical accuracy around 40% of investment) applied when only limited information on the process is available; _ study-type estimates, based on process flowsheets and global design data of the main process equipment (typical accuracy around 25%); and ~ detailed estimates, based on detailed engineering (typical accuracy around 5%-10%). In this study we are clearly confined to study-type methods of estimation. All methods in this category are based on the approach described by Lang [34]: the investment in a chemical plant is assumed to be the product of the equipment cost and a so-called Lang factor, which adds costs of installation, piping, instrumentation, plant construction, etc. : Investment = Lang factor X equipment costs (43)

An estimate of total investment is thus divided into the derivation of a Lang factor and an estimate of the equipment costs. The Lang factor Several authors derive Lang factors from cost analyses of existing chemical plants. These analyses vary between a global cost breakdown of a limited number of items [34] and more sophisticated detailed cost engineering

1351'
In Table 6 we compare four sources of Lang factor (inside battery limits) for our treatment plant, consisting of some 40% pressure vessels and columns, some 40% heat exchange equipment and around 20% for pumps. The J-ang factors coincide surprisingly well despite variations in basic assumptions. In this study the original Lang factor will be used adding 10% for the offsites,

TABLE 6. Comparison of Lang factors Guthrie

[ 3 51

Hirsh and Glazier [ 361 1.0

Lang [34]

Peters and Timmerhaus 1.0

[37]

Equipment

cost piping material

1.0 2.75 0.48 1.12 4.3.5

1.0 2.39 0.56 1.18 4.13 2.52 0.60 1.19 4.3 1

- direct plant costs excluding - piping material - overheads Total Lang factor (ISBL)* limits.

2.51 0.66 0.86 4.03

*ISBL = inside battery

19
TABLE 7. Example of investment in tray and trickle bed absorbers Basis: tray absorber (3.45 m X 18.5 m, 22 trays) and trickle bed absorber 7.05 m).

(1.75 m X 17 m, five sections

with a total packed

height of

Investments Tray 1. Pressure vessel (a) Vessel costs

in US k$ Trickle bed

[ 371
45 11 56

463 4 1 5

176

2. Internals and extra pumps (a) Tray/packing support material (b) Tray/packing support labour (c) Tray/packing support costs (d) Packing material including installation (e) Extra pumps including installation 3. Installation 3.1. Materials (a) Installation materials column 10 bar [ 37,401 (b) Tray/packing support installation materials (c) Installation materials costs at 10 bar Installation materials costs at 70 bar: 2.68 X item 3.1(c) 3.2. Labour (a) Labour in Europe (= 1.3 X US at 20 $/manhour [40]) (b) Overheads (80% of item 3.2(a)) (c) Total labour Sub-total 4. Contingency 5. Total installed and fees (45% of sub-total) column costs

+ 56

5 54 26

125 + 56 181 485 510 + 408 918 + 918 1922 + 865 2787

51 5 56 150

156 + 125 -281 + 281

692 + 311 1003

for exchanger type, the construction materials and the operating pressures [35]. The maximum exchange area is limited to 1000 m2 per unit. The capacity factor, being the product of flowrate (m3 s-) and head (bar), is representative of the costs of standard rotating equipment [37]. The high flowrate (0.25 m3 s-l ), high head (70 bar) booster pumps between the regenerator and the high pressure absorber do not fit in the standard manufacturing programs and have to be manufactured on request. The estimated cost, including drive and appendages, is US $ 130 000 per set. The working capital is calculated by: Working capital = chemicals + 0.5 x monthly Costingof hold-up (44)

salaries + 0.025 x investment

the treatment operation

The cost elements in gas treatment are: raw materials, energy, labour, maintenance, miscellaneous, depreciation and capital charges. 1. Raw materials comprise amine costs owing to pump losses, natural gas consumption in the TGU and catalyst replacement. The amine solvent hold-ups in the ARU and the TGU are estimated at 150 m3 each (absorber 35 m3, regenerator 40 m3, flash vessel 45 m3,piping, etc., 30 m3). Monthly pump losses are taken at 10% of total inventory. Annually, around 80 X lo3 kg of amine is consumed in the two units at a cost of US $0.90 kg- [l]. The natural gas consumption for the production of reducing gas is rated at US $ 0.13 NmP3. Catalyst life-

time in the SRU is reported to be 5 years [20]. For the reduction catalyst in the TGU the same lifetime is assumed. Thus an annual 20% of the catalyst inventory has to be replaced at US $ 2200/m for both the Claus and the TGU catalysts. The sulfur proceeds are credited at US $ 90/103 kg. 2. The energy costs involve electricity, steam, process and cooling water expenses. The electricity costs are rated at US $ O.O71/kW h. The treating process both consumes and produces steam in the regenerators and sulfur condensers respectively. The steam costs and proceeds are estimated at US $ 15/103 kg. The amount of process water used for steam make-up in the sulfur condensers is assumed to be 10% of steam consumed at a cost of US $ 2.00/m3. The cooling water costs are US $ 0.013/m3. 3. A total staff of 27 persons is assumed to involve an average salary cost of US $ 34000/man-year [41]. 5.5% is added for personnel charges and another 7% for tools, etc. 4. The cost of plant maintenance is set at 4% of the investments [37,42,43]. 5. Other costs include insurances, local taxes and disposal. Insurances plus local taxes are 2% of the investments [37,43] and waste disposal is estimated at 2% of the investments plus 2% of the costs of amine and catalyst consumed. The plant depreciates linearly over 10 years and capital charges are taken at 15Yo of investment plus working capital.

20

Results and Discussion

of Calculations

Several operating parameters in the flow schema of Fig. 1 (e.g. temperature and acid loading of the lean amine to the absorber, condensation temperature of the sulfur in the SRU) may be chosen at will within certain limits. By an overall optimization of the economics of the entire gas treatment plant, values for these operating parameters are assessed; an extensive optimization establishing all degrees of freedom is not envisaged. As an example, however, the influence of a partial economical optimization of the temperature approach of the lean, hot and rich, cold amine solutions at the regenerator outlets and inlets, respectively, is elaborated. Optimization of the exchanger temperature approach The parameter to be chosen freely is the temperature approach of the amine flows at the regenerator inlet and outlet. The smaller this temperature approach, the larger the amount of heat to be transferred and the smaller the average heat transfer driving force. Hence, the exchanger area becomes larger, as do the investments in the lean/rich exchanger. On the other hand, less steam is needed in the regenerator to balance the sensible heat difference between the entering and leaving streams (see eqn. (17)), so that smaller sized reboilers and regenerator columns can be installed. Obviously, an optimization of investments (exchange equipment plus regeneration) versus variable costs (steam) is at hand. For gas I and DIPA as the solvent at absorber conditions as in Table 2 a breakdown of the costs is given in Table 8 and shown graphically in Fig. 19. With increasing temperature approach the investment in the lean/rich exchanger decreases, as do the capital charges. On the other hand, steam consumption rises rapidly with an increasing temperature approach. The flat optimum of the total costs is found at a temperature approach of 8 C. It must be emphasized that, in
TABLE 8. Breakdown of costs vs. temperature approach (DIPA, gas I) Temperature 4 1. Investments (a) (b) (c) (d) (MS) exchanger 8.33 1.87 0.41 0.95 5.4 + 17.0 2. Energy requirements (a) Steam (kg s-t) (b) Electricity (kW h s-l) (c) Cooling water (m3 s-l) Annual energy costs (M$/year) Annual fmed costs (33% of investment) Total annual costs (M$/year) 16.66 0.70 0.52 8.17 5.61 14.38 17.54 0.70 0.55 9.16 4.65 13.80 18.44 0.70 0.54 9.54 4.22 13.76 19.44 0.70 0.62 10.00 3.93 13.93 20.40 0.70 0.65 10.43 3.83 14.26 21.35 0.70 0.68 10.85 3.70 14.55 + 14.1 5.42 1.92 0.42 0.98 5.4 + 12.8 3.96 1.98 0.43 1.01 5.4 + 11.9 3.02 2.03 0.45 1.03 5.4 + 11.6 2.39 2.09 0.46 1.21 5.4 + 11.2 1.99 2.14 0.48 1.24 5.4 approach 6 CC) 8 10 12 14

00
0

4 temperabum

8 approah

12 t T,

16

Pig. 19. Economical optimization loan/rich heat exchanger.

of the temperature

of the

principle, for each absorber-solvent-gas combination in Table 5 a different optimum temperature approach can be found. Considering the relative flatness of the optimum, however, we used the value of 8 C for all further evaluations. Comparison of the different combinations Eight absorber-solvent -gas combinations have been evaluated and compared with respect to the operating economics. The technical details are summarized in Table 5 and the costs breakdown is shown in Tables 9 t 10 and 11 + 12 for gas I and II respectively. Comparison of the investments in equipment for gas I and gas II (Tables 9 and 11 respectively) leads to the conclusion that the investments in plants with trickle bed absorbers are considerably lower than in the corresponding plants with tray absorbers. These investment

Lean/rich hat Regenerator Reboiler Amine cooler

(e) Absorber flash vessel, pumps compressor, condenser, reflux drum

(M$/year)

21
TABLE 9. Investment Amine Absorber type costs for agas 1 treatment plant (7% HsS and 10% COa) DIPA Tray Trickle 1.07 0.94 10.2 5.8 7.5 2.2 +28.9 25.7 +29.3 MDEA Tray 1.11 0.90 16.2 5.4 5.4 2.3 +23.1 Trickle 1.25 0.82 11.8 5.0 4.2 2.1

(a) Relative selectivity nrel (n/ntray, D~A, (b) Relative solvent circulation (@a/OQ. tray, Investments (M$) (1) Absorber regenerator unit (2) Sulfur recovery unit (Claus plant) (3) Tail-gas unit (SCOT) (4) Working capital

seeTable 5)
Alps, SW Table 5)

1 .OO 1 .oo 12.8 5.8 8.0 2.3 f-

Total investment

costs

TABLE 10. Treatment Amine Absorber type

costs of gas 1 in M$/year DIPA Tray DI~A) 1 .oo 1 .oo Trickle 1.07 0.94 1.11 0.90 MDEA Trickle 1.25 0.82

(a) Relative selectivity nrel (s/ntrav, (b) Relative solvent circulation 1. Raw materials 1 .l. Absorber-regenerator unit (a) Replacement amine 1.2. Sulfur recovery unit (a) Catalyst replacement (b) Sulfur produced 1.3. Tail-gas unit (a) Replacement reducing catalyst (b) Reducing agent (c) Replacement amine 2. Energy 2.1. Absorber-regenerator (a) Steam reboiler (b) Electricity (c) Cooling water 2.2. Sulfur recovery unit (a) Steam produced (b) Electricity (c) Process water 2.3. Tail-gas unit (a) Steam reboiler (b) Electricity (c) Cooling water 3. Labour 4. Maintenance 5. Other (a) Insurance, rents, local taxes (b) Waste removal/destruction 6. Depreciation 7. Capital charges Annual operating costs (MS/year) Cost of treating 1 Nm3 gas (+/Nm3) unit

0.04 0.41 -13.50 0.04 0.60 0.04

0.04 0.41

0.04

0.04 0.33 -13.50

_-13.50
0.04 0.59 0.04 0.03 0.54 0.04

0.03 0.49 0.04

7.95 1.4 1 0.22 -6.38 0.73 0.07 1.64 0.07 0.11 1.05 1.16 0.58 0.59 2.66 +4.34 3.83 0.24 +x

7.48 1.33 0.21 -6.34 0.73 0.07 1.62 0.06 0.10 1.05 1.03 0.51 0.52 2.35 +s 2.20 0.14

7.12 1.26 0.20 -6.35 1.06 0.07 0.88 0.04 0.08 1.05 1.17 0.59 0.60 2.70 +2.36 0.15

5.75 1.03 0.16 -6.19 1.06 0.07 0.58 0.01 0.06 1.05 0.92 0.46 0.47 2.10 3.47

-1.57 -0.10

reductions are caused by lower absorber costs in the ARU, and smaller SRLJs and TGUs owing to the increased

temperatures

required

for

the

MDEA

regeneration

and

selectivities. The investments for MDEA ARUs are higher than for their DIPA counterparts because of the higher

the consequently considerably larger lean/rich exchangers. The treatment costs are summarized for eight combinations in Tables 10 and 12 for gas I and II respectively.

22
TABLE 11. Investment Amine Absorber type costs for a gas II treatment plant (5% H2S and 10% CO2) DIPA Tray 1 .oo 1 .oo 13.3 4.8 10.8 +2.3 31.2 Trickle 1.11 0.92 10.5 4.7 9.5 +2.2 26.9 MDEA Tray 1.14 0.88 16.8 4.3 7.0 +2.3 30.4 Trickle 1.40 0.72 11.4 3.9 4.4 +2.1 21.8

(a) Relative selectivity qreI (b) Relative solvent circulation Investments (M$) (1) Absorber-regenerator unit (2) Sulfur recovery unit (Claus plant) (3) Tail-gas unit (SCOT) (4) Working capital Total investment costs

TABLE 12. Annual Amine Absorber type

treatment

costs for gas 11 in M$/year DIPA Tray 1 .oo 1 .oo Trickle 1.11 0.92 MDEA Tray 1.14 0.88 Trickle 1.40 0.72

(a) Relative selectivity qrel (b) Relative solvent circulation 1. Raw materials 1.1. Absorber-regenerator unit (a) Replacement amine 1.2. Sulfur recovery unit (a) Catalyst replacement (b) Sulfur produced 1.3. Tail-gas unit (a) Reducing agent (b) Others 2. Energy 2.1. Absorber-regenerator (a) Steam reboller (b) Others 2.2. Sulfur recovery unit (a) Steam produced (b) Others 2.3. Tail-gas unit (a) Steam reboiler (b) Others 3. Labour 4. Maintenance 5. Other 6. Depreciation 7. Capital charge Annual operating costs (MS/year) Cost of treating 1 Nm3 gas ($/Nm3) +* unit

0.04 0.3 5 -9.72 0.52 0.06

0.04 0.33 -9.72

0.49

0.04
0.27 -9.12

0.04 0.25 -9.72 0.37 0.06

0.06

0.45 0.06

8.50 1.73 -4.49 0.57 2.67 0.24 1.05 1.25 1.25 2.89

7.77 1.58 -4.41 0.57 2.50 0.21 1.05 1.08 1.09 2.47 +4.04 9.15 0.58 +s

7.48 1.52 -4.43 0.81 1.56 0.16 1.05 1.22 1.23 2.81 +9.07 0.58

5.46 1.13 -4.25 0.81 0.71 0.09 1.05 0.87 0.89 1.97 3.27 3.00 0.19

11.59 0.74

In fact three costs items are overruling in each case: energy (in particular steam) costs, sulfur proceeds and depreciation + capital charges. The treatment costs per Nm3 of natural gas are shown as a function of the relative selectivity rjrel (= v/r)tray, DIPA) in Fig. 20. Two very important conclusions can be drawn from this Figure. Firstly, the treatment costs at a fixed feed gas composition decrease considerably with increasing selectivity in the high pressure absorber. These costs can even be converted into a net income at higher relative

selectivities: qrel > 1.2 for gas I and qrel > 1.55 for gas II. Secondly, the more H2S in the feed gas (H2S, gas I > H2S, gas II), the lower the treatment costs. This effect is mainly due to higher sulfur proceeds and lower overall energy costs in treating the H2S rich gas. The latter is illustrated in Fig. 21, where the energy costs for the ARU, SRU, and TGU are plotted as a function of the relative selectivity. The ARU energy costs for gas 11 are 5%-10% higher than for gas I at equal vrel because in treating gas II roughly 5%-10% more acid gases are

23
0 OIPA. tray LI mlEA.tray l DIPA,tickle-bed 9 WOEA. trickle-bed 07% 0 \
IO

treat\ng costs per Nm3 IUS $c/Nm3l

nnual operahng costs IUS Ml/al 5

.\
10 11 relative 12 13 relectivtty. a,,( 14 15

Fig. 20. Treatment costs per Nm3 of natural gas as a function of the relative selectivities.

sulfur price is established by a market mechanism, the proceeds will fluctuate and will have a varying beneficial contribution to the treatment costs:The influence of the sulfur price on the treatment costs is shown in Fig. 22. At increasingly higher sulfur prices the treatment processes will eventually become profitable, the HaS rich gas 1 being more dependent on the sulfur price. In Fig. 23 the treatment costs are shown as a function of the relative energy costs, defined as energy costs/ energy costs used in this study. These relative energy costs evidently exercise a large influence on the energy intensive, low selectivity process operations, e.g. treating gas II in the tray absorber using DIPA as the solvent (see also Fig. 21). In the high selectivity operations using trickle bed absorbers and MDEA as the solvent, the energy consumption and thus the impact of energy costs on the treatment economics is relatively low. This is an extra incentive for using more selective high pressure absorption equipment, such as the cascade of trickle bed absorbers.

0 OIFn, tray 0 tlOEA.tray . OIW. tnckle-bed n tlDEA.hckle-bed --oar

-10

.-.- --..-

wA.tray MDEA.tray OIW, trickle- bed PDEA.tr,ck,e-bed 20

annual energy

costs

OJS Mb)

--a

costs 3 E EN&,
0

treatifq

-0.2

-O~M-(

1
-o-.1.0

_-O____m_.-_-_-----.-SRLI .3-=---5 -10

VI

Ire, Fig. 21. Energy costs for treating gas I and gas II.

13 I.2 relative selectlvl~,

IL

15

i
sulphur

-1 loo

2b (US

price

$/ldkQl

Fig.22. Influence of sulfur price on treatment costs.

absorbed since gas II contains 15% and gas 1 only 14% of H2S plus CO2 (see Table 5). In the SRU, more steam is produced in processing gas 1 because of the higher amounts of H$ converted. Hence the SRU steam generation is larger for gas 1. After the reduction of unconverted SO2 to H2S, the HsS concentration to the TGU absorber is higher for gas I than for gas II and higher selectivities are thus obtained. This effect results in lower TGU solvent flowrates and therefore reduced steam requirements for gas I in the TGU regenerator. The overall energy picture for ARU + SRU + TGU results in energy costs which are -3 M$/year, or 0.2 US Q/Nm3 of treated gas, lower for gas I. Combination of these costs with the difference in sulfur proceeds, ie. 3.78 M$/year or 0.24 $/Nm3 of treated gas, brings the total difference between the treatment costs of gas 1 and II to 6.8 M$/year, which is equivalent to 0.44 $/Nn-? of treated gas. Sulfur proceeds have a considerable impact on the treatment economics (see Tables 10 and 12). Since the

DIPA, tray

.-.MDEA.tray - - - DIPA. twkle-bed -. - HLXA, trickle-bed

20

5 total annual costs IUS M$/al

IO

I
10

relahve energy
Fig. 23. Treatment costs. costs

1.5

20

costs

as a function

of the relative energy

24

Conclusions
Two models have been developed which describe the reaction and equilibrium combined mass transfer, processes in the absorption of HsS and CO? by aqueous amine solutions in tray and cocurrent trickle bed absorbers respectively. The analytical and numerical solutions of the two mass transfer models used in these absorber calculations give almost equal results despite their essentially different principal assumptions and complexity. The impact of variations in design and operating parameters on the tray absorber performance has been established. The operating pressure of the absorber exercises a negligible influence on the absorber selectivity and the solvent rate. The acid gas loadings of lean and rich amine solvents and the total amine concentration have a significant effect on selectivities and solvent flowrates. A cascade of trickle bed reactors produces higher H2S absorption selectivities than tray absorbers at identical operating and design conditions. This effect is mainly due to the increased kg/kp ratio realized in trickle bed reactors. In the evaluation of the economics of the use of eight absorber-solvent-gas combinations it is shown that the investments for the trickle bed absorbers are considerably lower than for their tray absorber counterparts. Moreover, as a result of the increased selectivity and lower investment in the trickle bed absorbers, the annual operating costs for the treatment plant are considerably lower with this absorber type. The use of MDEA solutions is economically more attractive than of DIPA solutions.

Rev R overall, liquid and gas phase mass transfer Rg resistances, defined by eqn. (32) s m-l Re = d,vp/p, Reynolds number SC = pl/Dp, Schmidt number T temperature, K V superficial velocity, m s-r X liquid phase molar fraction gas phase volume fraction concentration, mol rnp3 acid gas loading, defined by eqn. (6) selectivity defined by eqn. (33) viscosity, kg m-l s-l density, kg mW3 packing constant in eqns. (28) and (29) gas flowrate to tray i, Nm3 s-l solvent flowrate m3 s-l steam flowrate, kg s-l steam flowrate, kg (m3 solvent)- Indices FIL !2 z st tot gas phase heat limited liquid phase overall stripping limited steam total (in reacted and unreacted

form)

References
A. L. Kohl and F. C. Riesenfeld, Gas Purification, Gulf Publ. Co., Houston, TX, 1979. R. N. Maddox, Gasand Liquid Sweetening, John M. Campbell, Norman, OK, 1977. M. W. McEwan and R. C. Darton, J. Sep. Process Technol., 1 (1980) 1. M. W. McEwan and A. Marmin, Developments in LNG treating, Froc. 6th Inr. ConfI on Liquefied Natural Gas, Kyoto, Japan, April 1980, Vol. 2, Paper 3, Appendix. C. Ouwerkerk, Hydrocarbon Process., 57 (1978) 89. C. Blanc, J. Elgue and F. Lallemand, Hydrocarbon Process., 60 (1981) 8,111. C. Ouwerkerk, Process selection for the treating of natural gas, Insr. Chem. Eng. Symp. Ser., 44 (1) (1976) 36. A. E. Cornet&en, Tmns. Inst. Chem. Eng., 58 (1980) 242. P. M. M. Blauwhoff, Thesis, Twente University of Technology, The Netherlands, 1982. P. M. M. Blauwhoff, G. F. Versteeg and W. P. M. van Swaaij, Chem Eng. Sci.. 39 (1984) 207. P. M. M. Blauwhoff and W. P. M. van Swaaij, Proc. 2nd Znt. Conf Phase Equilibria Fluid Properties in the Chemical Industry, Berlin, March 1980, Dechema, Frankfurt/Main, 1980, EFCE Publ. Ser. 11, Part I, p. 78. P. M. M. Blauwhoff and M. Bos, J. Chem. Eng. DQ~Q, 26 (1981) 7. R. Cornelisse, A. A. C. M. Beenackers, F. P. H. van Beckum and W. P. M. van Swaaij, Chem. Eng. Sci.. 35 (1980) 1245. P. V. Danckwerts and M. M. Sharma, Chem. Eng. (London), 10 (1966) 244. G. Soave, Chem. Eng. ScL, 27(1972) 1197.

Nomenclature
A a

tray
,,

:I
c2 CP

d, d, &
fl

AH

AH,,
He
J

k M m ; R

gas-liquid interfacial area on tray, m2 specific gas-liquid interfacial area, m2/m3 gas-liquid interfacial area, m2/m2 tray constant in eqn. (26) constant in eqn. (27) specific heat, kJ kg- C- diffusion coefficient m2 s- particle diameter, m column diameter, m equivalent spherical packing diameter [32], m enhancement factor equilibrium approach factor, defined by eqn. (IS) heat of reaction + absorption, kJ mol- latent heat of steam, kJ mall Henrys coefficient, mol rn- bar- molar flux, mol rn- s-l mass transfer coefficient, m s-r molar mass, kg rnol- dimensionless solubility, defined by eqn. (16) total number of trays pressure, bar ratio, defined by eqn. (19)

5 6 I 8 9 10 11

12 13 14 15

25
16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 R. M. Secor and J. A. Beutler, AIChE J., 13 (1967) 365. II. Hikita and S. Asai, Kagaku Kogaku, 27 (1963) 823. D. R. Olander, AZChE J., 6 (1960) 233. B. W. Gamson and R. H. Elkins, Chem. Eng. Prog.. 49 (1953) 203. H. Fisher, Chem.-lug.-Tech., 39 (1967) 515. R. E. Treybal, Mass Transfer Operations, McGraw-Hill, New York, 3rd edn., 1980. B. C. Goar, Sulfur recovery from sour natural gas, Energy Commun., 2 (5) (1976) 475-520. W. M. Kays and A. L. London, Compact Heat Exchangers, McGraw-HlIl, New York, 1964. T. J. Edwards, G. Maurer, J. Newman and J. M. Prausnitz, AIChE J., 24 (1978) 966. K. Schwabe, W. Graichen and D. Spiethoff, Z. Physik. Chem. (Frankfurt amMain), 20 (19.59) 68. R. C. Reid, J. M. Prausnitz and T. K. Sherwood, The Properties of Gases and Liquids, McGraw-Hill, New York, 1977. A. Tavares da Silva and P. V. Danckwerts, Inst. Chem. Eng. Symp. Ser., 28 (1968) 48. M. M. Sharma and R. K. Gupta, Trans. Inst. Chem. Eng., 45 (1967) 169. G. Nonhebel, Gas Purification Processes for Air Pollution Control, Newnes-Butterworths, London, 1972. S. Fukushlma and K. Kusaka, J. Chem. Eng. Jpn., 10 (1977) 461. 31 S. Fukushima and K. Kusaka, J. Chem. Eng. Jpn., 10 (1977) 468. 32 S. Pukushima and K. Kusaka, J. Chem. Eng. Jpn., 11 (1978) 241. 33 T. Shridhar and 0. E. Potter, Znd. Eng. Chem. Fundam., 19 (1980) 21. 34 H. J. Lang, Chem. Eng., 54 (10) (1947) 117; 55 (6) (1948) 112. 35 K. M. Guthrie, Capital cost estimating, Chem Eng., 76 (Mar. 24) (1969) 114. 36 J. Il. Hirsh and E. M. Glazier, Estimating plant investment costs, Chem. Eng. Prog., 56 (1960) 37. 37 M. S. Peters and K. D. Timmerhaus, Plant Design and Economics for Chemical Engineers, McGraw-Hill, New York, 1980. 38 A. Mullet et aZ., Estimate costs of pressure vessels via correlations, Chem. Eng., (Oct. 5) (1981) 145. 39 A. Pikulic and H. E. Diaz, Costs of process equipment and other items, Chem. Eng., 84 (Oct. 10) (1977) 106. 40 J. S. Miller and W. A. Kapella, Installed costs of a distillation column, Chem. Eng., (Apr. 11) (1977) 124. 41 CBS, Zakboek 1980, Den Haag, The Netherlands, 1981. 42 J. Happel, Chemical Process Economics, Wiley, New York, 1958. 43 F. C. Jelen, Cost and Optimization Engineering, McGraw-Hill, New York, 1970.